925 
um aesth’uni L.) plus 40 
Fluorescence of the soils, 
sidues and soil remained 
iminated from 35 of the 
luded that fluorescence 
[ues. 
Advances in low light 
fluorescence from crop 
ling as to the amount of 
tive was to demonstrate 
ybion Electronics Corp.. 
isified camera assembly 
th the video scan rate to 
n to provide continuous 
n the locked filter mode 
i was not necessary' and 
о images were grabbed 
l, 535-570 nm. 650-685 
itimal for measuring the 
г this proof-of-concept 
nultis pec trai camera 
d soybean residue, 
iminated the scene. 
;y any endorsement of 
Table 1 Reflectance factors and relative fluorescence intensities (RFI) for two soils and the soybean residue. 
Daughtry et al. (1993). 
SAMPLE 
MOISTURE 
TM1 1 
TM2 
TM3 
TM4 
RFI 2 
Barnes 
Dry 
0.07 
0.09 
0.13 
0.21 
10.1 
Wet 
0.03 
0.04 
0.06 
0.09 
5.4 
Codorus 
Dry 
0.15 
0.21 
0.27 
0.33 
17.7 
Wet 
0.05 
0.08 
0.11 
0.15 
7.1 
Soybean Residue 
Dry 
0.15 
0.18 
0.22 
0.29 
54.0 
Wet 
- 
- 
- 
- 
42.8 
"ГМ1 - 450-520 nm; TM2 - 5200-600 nm; TM3 - 630-690 nm; TM4 - 760-900 nm. 
2 Relative Fluorescence Intensity in 420-550 nm wavelength band. 
Pixels with values >6 were classified as soybean residue and were designated as white in Figure 5. The 
remaining pixels with a value < 6 were classified as soil and were designated as black in Figure 5. Using this simple 
two class scheme, we identified 23.2% of the pixels as residue which is close to the measured 22.2% residue cover. 
Inspection of the classification image clearly reveals that most of the residue is accurately represented there are 
some noticeable errors of omission, where portions of a continuous stem are classified as soil, as well as some errors 
of commission, where small area of soil are classified as residue. During a post-classification of the scene, we 
noticed seme small pieces of plant material in the soil presumably remaining from the previous crop. Therefore the 
actual residue cover was slightly greater than expected 
Table 2 shows the classification results for both soils. Moisture quenched the fluorescence of both the soil 
and residue, as Daughtry et al. (1993) reported, but had little effect cm classification accuracy. The results were 
similar for both soils. 
Table 2. Classification results of fluorescence images of soybean residue on two soils. The actual soybean residue 
cover was 22.2%. 
SOIL 
MOISTURE 
COVER, % 
ERROR 1 , % 
Barnes 
Dry 
23.9 
1.7 
Wet 
24.8 
2.6 
Codorus 
Dry 
23.2 
1.0 
Wet 
21.3 
-0.9 
Error - Cover estimated by fluorescence minus measured cover. 
The threshold can be approximated by iteratively classifying and viewing the images until most of the pieces 
of residue are correctly identified. In our experience, as one approaches the correct classification threshold, the 
percent residue cover changes very little. 
In conclusion, we have demonstrated that residue cover can be determined using video immaging of crop 
residue fluorescence. Furthermore, fluorescence techniques are less ambiguous and better suited far discriminating 
crop residues than reflectance methods. The video images can provide a permanent record of the percent cover 
conditions in a field and can be reanalyzed as needed. Video imaging of crop residue fluorescence provides a 
intuitive understanding of the amount of residue cover as compared to non-imaging techniques. Additional 
development and testing is still required to produce a portable imaging system capable of quantifying crop residue 
cover in the field